ME News reports that on July 22 (UTC+8), the Xiao Hong Shu technology team open-sourced BigMac, a new paradigm for dependency-safe nested pipelines tailored for multimodal large model training. Built around an LLM pipeline, BigMac integrates encoder and generator computations without disrupting execution order, achieving 1.08x to 1.9x speedup over baseline implementations while maintaining bounded activation memory. BigMac is already in production as a core component of the dots multimodal model training pipeline. 🔗 Read the original article: https://mp.weixin.qq.com/s/tNJARtn1jIayL5URg87Row via AI HOT · https://aihot.virxact.com/items/cmrvx8ycy0240bipz0pkrvryq (Source: AiHot)
Xiaohongshu open-sources BigMac, a new multimodal large model training paradigm.
KuCoinFlashShare
Summary
Xiaohongshu open-sources BigMac, a new multimodal training paradigm built on MetaEra. The model employs a dependency-safe nested pipeline to integrate encoder and generator computations within an LLM pipeline, achieving a 1.08x to 1.9x speedup over standard methods while maintaining bounded activation memory. BigMac now supports production training for the Dots multimodal model. Open interest in technical indicators continues to grow as the model gains adoption.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.